AI-produced proofs create rival publishing models for mathematical research
Two AI labs are testing different ways to turn machine-generated mathematics into knowledge that people can verify and use.
OpenAI released hundreds of claimed mathematical breakthroughs in a bulk public dump, some with Lean certificates, while the source says humans had yet to understand most of the proofs. Anthropic instead compensated selected mathematicians to digest and announce an AI-derived result, placing a private lab in charge of choosing the human interpreters. The contrast turns verification, exposition, credit and payment into new publishing functions as AI labs become producers and distributors of scholarly knowledge.